Loci Impacting Polymorphic Gait in the Tennessee Walking Horse1

Loci impacting polymorphic gait in the Tennessee Walking Horse1

E. A. Staiger,* M. A. Abri,† C. A. S. Silva,‡ and S. A. Brooks§2

*Department of Animal Science, Cornell University, Ithaca, NY 14853; †Department of Animal and Veterinary Sciences, College of Agriculture and Marine Sciences, Sultan Qaboos University, PO box 34 Al Khod, Postal Code 123, Muscat, Oman; ‡Department of Animal Science, Instituto Federal Sul-Rio-Grandense, Pelotas, Rio Grande do Sul 96060-290, Brazil; and §Department of Animal Science, University of Florida, Gainesville 32611 Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 ABSTRACT: Following domestication, man select- 129 TWH and genotyping was performed at approxi- ed the horse primarily for the purpose of transporta- mately 60,000 loci using the Illumina Equine SNP70 tion rather than consumption; this selective strategy beadchip at GeneSeek Inc. (Lincoln, NE). Case–con- created divergent traits for locomotion. At intermedi- trol association tests identified suggestive regions for ate speeds, beyond the flat walk, the horse can per- gait type on equine chromosome (ECA) 19 (P-value form a range of diagonal and lateral 2-beat or 4-beat of 1.50 × 10–5 after 1 million permutations; PLINK gait patterns. The Tennessee Walking Horse (TWH) version 1.07). Haplotype analysis identified 2 sig- is the only U.S. breed able to perform an even-timed nificant haplotypes on ECA19 and ECA11 (P-values 4-beat gait (the “running-walk”) at intermediate of 3.7 × 10–5 and 3.92 × 10–5, respectively). Genes speeds; however, within the breed, there is remaining within these suggestive regions play roles in develop- variation in gait type. To investigate the contribution mental processes and biological regulation, indicat- of genetics to this unique trait, blood or hair samples ing there may be variant differences in the neurobiol- for DNA and gait information were collected from ogy and regulation of horses with a polymorphic gait. Key words: gait, genome-wide association study, horse, running walk

INTRODUCTION out a moment of suspension is commonly called a “gaited” horse. The horse is a unique model for the Horses are polymorphic in the pattern and tim- study of gait development, as no other mammalian ing of locomotion. At intermediate speeds, the horse species is known to discretely segregate for con- can perform a range of diagonal and lateral 2-beat or genital differences in stride cadence and footfall pat- 4-beat footfall patterns (Harris, 1993). A horse with tern. Laboratory-induced models of variation in gait the ability to perform any of the 4-beat gaits with- exist in mice and cats via transverse sectioning of the spinal cord (Grillner and Zangger, 1979) and in- 1This work was supported in part by a seed grant from the duced mutations or knockouts in mice (Coonan et al., Cornell Center for Vertebrate Genomics, by the USDA National 2001; Kullander et al., 2001; Hinckley et al., 2005; Institute of Food and Agriculture, Hatch project NYC-127454, and Wilson et al., 2005; Crone et al., 2008). These labo- by a grant from the Foundation for the Advancement and Support ratory models enabled discovery of localized neural of the Tennessee Walking Show Horse. Any opinions, findings, networks known as central pattern generators (CPG; conclusions, or recommendations expressed in this publication are Kuo, 2002). When locomotion is initiated, activity those of the authors and do not necessarily reflect the view of the in the CPG is turned on and maintained by inputs National Institute of Food and Agriculture (NIFA) or the United from descending locomotor commands originating States Department of Agriculture (USDA). We appreciate the help from neurons in the brainstem and midbrain (Kiehn, of all contributing owners, breeders, and trainers. 2006). Naturally occurring models of lateral gait, 2Corresponding author: [email protected] Received October 2, 2015. such as horses, provide a novel resource for study of Accepted January 29, 2016. CPG structure and function. 1377 1378 Staiger et al. Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021

Figure 1. Two foals that exemplify variation in gait types in Tennessee Walking horses. (A) The diagonal trot of a multigaited horse (foal is 2 mo old) and (B) the lateral amble of a lateral-only horse (foal is 2 d old).

The Tennessee Walking Horse (TWH) is a U.S. travel, approximately at the horse’s wither height, and breed renowned for their ability to perform an even- far enough away that the horse fit within the camera’s timed 4-beat gait (the “running-walk”) at intermedi- view finder. We recorded each horse for at least 10 s at ate speeds. The TWH is nearly fixed for the DMRT3 the walk and all intermediate gaits. We also collected mutation (Promerová et al., 2014). However, TWH level of training under saddle, horse discipline (breed- are capable of performing the whole range of inter- ing, trail, and show), type of shoes (barefoot, alumi- mediate gaits from 2-beat lateral to 4-beat to 2-beat num, steel, and pad), and pedigree information. diagonal (Tennessee Walking Horse Breeders’ and A single experienced observer classified horses Exhibitors’ Association, 2011). At liberty, these horses as either lateral only or multigaited (able to perform show an innate preference for either trot, pace, or run- both lateral and diagonal gaits, including the trot) after ning walk from birth (Fig. 1). The objective of this review of video recordings in slow motion. The in- study is to identify candidate loci and SNPs unique to dividual classified horses as lateral-only gaited if the gait type variation within the TWH breed through a lateral pair of legs traveled together in a 4-beat pat- genome-wide association study (GWAS). tern (with the hind hoof landing before the front hoof). If the footfall landing followed a 2-beat or 4-beat di- MATERIALS AND METHODS agonal pattern and the lateral gait pattern previously described was also observed, the individual classified All animal procedures were approved by the the horse as multigaited. Horses were also classified Cornell University Institutional Animal Care and Use as multigaited if the owner reported viewing the horse Committee (protocol 2008-0121). trotting at liberty but not in the video analysis (n = 9). Eighty-two horses were classified as lateral only and Animals and Phenotypes 57 as multigaited in a case–control GWAS design. Pedigree information, as contained in the reg- We collected 139 TWH samples from 55 private istry or reported by the owner, was recorded for 137 farms and public horse shows across the United States horses; the remaining 2 horses were registered, but the during a period from 2010 to 2015 for use in this study. registration papers including the pedigree had been The group consisted of 82 mares and 57 geldings and lost. Horses were selected so that no horse was a first- stallions ranging from 1 to 34 yr in age at time of col- generation relative to any other within a given phe- lection. We video recorded all horses either under sad- notypic category to reduce population stratification. dle, at liberty in a round pen, or in hand at the walk and Half-siblings (n = 17 pairs) or parent–offspring pairs at all intermediate gaits using a Canon FS30 camcord- (n = 8 pairs) were only used if 1 member of each pair er (Canon Inc., Tokyo, Japan). The camcorder was set had been classified as lateral-only gaited and the other up so that it was perpendicular to the horse’s plane of as multigaited to balance any stratification across the Genetics of gait in Tennessee Walking horses 1379

2 gait phenotypes. Pedigraph version 2.4 (Garbe and TWH randomly selected from this study to analyze Da, 2008) and the 5-generation pedigrees were used population structure. Values were binned in groups to evaluate the inbreeding within the 137 horses. In of 5,000 consecutive markers, and average r2 and the 2 cases where pedigree information was not avail- inter-SNP distance were graphed using Excel 2007 able, we calculated genome-sharing ( pi ) values and (Microsoft Corp., New York, NY). inbreeding (F) for the autosomes across all 139 individuals in PLINK (Purcell et al., 2007) after pruning DMRT3 Amplification and Genotyping for minor allele frequency (MAF) of 0.05 and genotyping rate of 0.05. A recent publication describes a mutation in the DMRT3 gene, which is suggested to control for the Deoxyribonucleic Acid Extraction, Genotyping, ability for a horse to perform lateral patterned gaits and Quality Control (Andersson et al., 2012). We examined the effect of this mutation in our study population, following a Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 We isolated genomic DNA from blood samples using previously published protocol developed in our lab the Gentra Puregene Blood Kit, following the manufac- (Patterson et al., 2015). turer’s protocol for whole blood (Qiagen Inc., Valencia, CA). We performed extraction of DNA from hair using Statistical Analysis the Gentra Puregene DNA Isolation Kit, following the manufacturer’s protocol with modifications to optimize PLINK version 1.07 (Purcell et al., 2007) and SVS for hair root bulbs (Qiagen Inc.; Cook et al., 2010). software from Golden Helix, Inc. (http://goldenhelix. We genotyped 139 TWH with complete gait phe- com; accessed December 5, 2014) were both used to notypes at 65,157 loci using the Equine SNP70K bead- test the resulting 42,226 genotypes in 129 individuals chip (Illumina Inc., San Diego, CA) at GeneSeek Inc. for significant association with gait type. We applied (Lincoln, NE) and added 2 additional SNP individually case–control allelic association and basic adaptive per- genotyped as described below. Genome-wide identity- mutation (–perm in PLINK) to the data set, with later- by-descent estimates were calculated using an identity- al-only individuals run as controls and multigaited as by-state similarity matrix to evaluate population struc- cases due to the higher prevalence of lateral-only horses ture after filtering for MAF and genotyping rate as de- in our sampling population. Multidimensional scaling scribed above in PLINK (Purcell et al., 2007). Twelve was performed using PLINK version 1.07 on all SNP pairs (but only 22 individuals) had identity-by-state that passed quality control filters using the “–mds-plot similarities greater than 0.80; only 1 pair was greater 4” option. To assess association to haplotypes, candidate than 0.90 and the remaining 10 pairs were less than regions of 2 Mb centered on the most significant SNP 0.85. We removed 5 individuals from these pairs: 1 in- from equine chromosome (ECA) 11, ECA12, ECA19, dividual per pair if the pair was of the same gait type or and ECA25 were selected and filtered to include only if an individual appeared in 2 or more pairwise compar- markers with a 99% genotyping rate and MAF > 0.01. isons. Single nucleotide polymorphisms with less than We tested for a haplotype association to gait type by 90% genotyping rate (n = 10,035) and with a MAF < a logistic model, permutation, and χ2 test using slid- 0.05 (n = 15,417) were excluded. Five individuals were ing windows of between 2 and 20 SNP in PLINK. The excluded for a genotyping rate < 86%. The final data set LD structure between markers was examined using consisted of 129 individuals (52 cases and 77 controls) Haploview version 4.2 (Barrett et al., 2005). Epistasis genotyped at 42,226 markers with a mean call rate of between candidate loci and all other SNP were tested 96.56%. We used a Bonferroni significance cutoff of in PLINK (–epistasis). A Student’s t test was applied 1.32 × 10–6, conservatively estimating 37,770 indepen- to compare age differences between cases and controls. dent comparisons following removal of 4,456 markers Results were visualized using the JMP Pro version 11 that are in complete linkage disequilibrium (LD; with software package (SAS Inst. Inc., Cary, NC). an r2 > 0.99) from our 42,226 total markers in the study. Genomewide LD was estimated using the r2 sta- RESULTS tistic in PLINK (Purcell et al., 2007) under the following filters: MAF < 0.05 and deviation from Hardy– Genome-wide Association Study Weinberg equilibrium P < 0.0001. Ten individuals Identifies a Suggestive Candidate Locus previously genotyped on the Equine SNP50 beadchip for Gait Type in the Tennessee Walking Horse (Illumina Inc., San Diego, CA) were chosen from the Arabian, Egyptian Arabian, Thoroughbred, and Genotypes generated for 129 horses on the Equine Saddlebred breeds and compared with 10 unrelated SNP70K beadchip enabled a GWAS to identify loci con- 1380 Staiger et al.

Table 1. Top 14 Tennessee Walking Horse genome-wide association SNP associated with gait type

Position, Major Minor Raw Permutated No. of SNP Chromosome bp allele allele MAF1 P-value P-values permutations 19 BIEC2_437782 37,674,807 G A 0.30 1.71 × 10–5 1.50 × 10–5 1,000,000 19 BIEC2_437781 37,674,757 C T 0.33 4.12 × 10–5 6.06 × 10–5 594,405 12 BIEC2_175614 12,173,695 T C 0.09 4.60 × 10–5 7.08 × 10–5 508,491 25 BIEC2_659946 12,939,295 C A 0.48 5.58 × 10–5 7.83 × 10–5 460,000 11 BIEC2_156863 47,974,247 G A 0.12 1.15 × 10–4 7.89 × 10–5 456,543 19 BIEC2_456655 30,903,484 A C 0.23 1.34 × 10–4 4.17 × 10–4 88,822 8 BIEC2_1037747 24,686,529 G A 0.14 1.46 × 10–4 2.25 × 10–4 160,195 25 BIEC2_658647 10,385,019 A C 0.23 2.23 × 10–4 3.83 × 10–4 93,906 13 BIEC2_210428 11,381,404 T C 0.40 2.39 × 10–4 1.74 × 10–4 207,376 19 BIEC2_428949 11,617,833 A G 0.06 3.11 × 10–4 2.60 × 10–4 138,583 Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 10 BIEC2_124370 53,570,894 T G 0.48 3.70 × 10–4 6.36 × 10–4 56,601 19 BIEC2_429943 13,181,184 A G 0.30 5.20 × 10–4 1.00 × 10–3 36,000 19 BIEC2_452486 16,010,941 A G 0.43 1.16 × 10–3 2.64 × 10–4 136,178 19 BIEC2_430358 16,034,924 T G 0.43 1.16 × 10–3 2.64 × 10–4 136,178 1MAF = minor allele frequency. tributing to gait type variation within TWH. Allelic as- dred thirty-eight horses had the AA genotype and 1 sociation identified several candidate markers on ECA19, horse had the CA genotype (frequency of A = 0.9964), ECA12, ECA25, ECA11, and ECA8 with raw P-values indicating that DMRT3 is not a predictor for gait type ranging from 1.71 × 10–5 to 1.46 × 10–4 (Table 1); how- within TWH (P = 0.2227). Age was not different be- ever, none surpassed a Bonferroni correction for mul- tween lateral-only and multigaited horses (P = 0.13). tiple testing. Application of a permutation test supported The TWH is recognized as a breed with mod- the top candidate region from the allelic association on erate to high levels of inbreeding due to common ECA19 (Fig. 2a) with a P-value of 1.50 × 10–5 (Table 1). line-breeding practices early in breed development The 2 top candidate markers on ECA19 are (Fletcher, 1946), which can also be observed in the in complete LD and span just 51 bp within an ap- 5-generation pedigree of horses used in this study proximately 2-kb intron of FBX040 (Fig. 2e and 2f). (Supplemental Fig. S1a; see the online version of Additional significant markers on ECA19 are approxi- the article at http://journalofanimalscience.org). The mately 26 Mb downstream from this region, spanning available genomewide genotypes and pedigrees al- 23 kb in a 967-kb region that did not contain any an- lowed us to further examine what, if any, substructure notated genes. The associated marker on ECA12 falls within the breed may be influencing gait type prefer- within a 220-kb region lacking any annotated features. ence within individuals, as observed in the Icelandic The marker on ECA25 falls within a 207-kb region horse (Albertsdóttir et al., 2008). Genomic inflation also without known features. The marker on ECA11 was limited to a factor of 1, indicating adequate con- falls within a 315-bp intron of GGT6 (Fig. 2b and 2c). trol of population stratification. Further examination There was no evidence for epistatic interaction be- of the multidimensional scaling (MDS) analysis did tween the top allelic association markers from ECA19, not identify significant population stratification due to ECA25, ECA11, and ECA12. gait type in the TWH (Supplemental Fig. S1c and S1d; Haplotype analysis revealed associations of gait see the online version of the article at http://journalo- type with a 3-SNP block spanning ECA19, 37,674,757 fanimalscience.org). Two subpopulations are detect- to 37,719,196 (alleles TAG; P = 3.79 × 10–5, able in the comparison of dimensions 1 and 3, but Bonferroni cutoff of 6.25 × 10–4), and a 7-SNP block these are not correlated with gait type (Fig. 3a). The spanning ECA11, 47,893,127 to 48,015,457 (allele identified subpopulations are likely due to popular sire GCGAGGA; P = 3.92 × 10–5, Bonferroni cutoff of effects as observed in the pedigree tree (Supplemental 5.95 × 10–4). The ECA19 SNP block genomic region Fig. S1b; see the online version of the article at http:// includes the FBXO40, ARGFX, and POLQ genes (Fig. journalofanimalscience.org) and could explain the 2f and 2g). The ECA11 SNP block genomic region slight skew observed in the quantile–quantile plot includes the SPNS3, SPNS2, MYBBP1A, GGT6, and (Supplemental Fig. S2; see the online version of the SMTNL2 genes (Fig. 2c and 2d). article at http://journalofanimalscience.org). Pedigree- All horses in this study were genotyped for the inferred inbreeding coefficients were moderate in both DMRT3 mutation (Andersson et al., 2012). One hun- lateral-only and multigaited horses, 7.9 (range 0.32 to Genetics of gait in Tennessee Walking horses 1381 P -values.

10 Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 Figure Figure 2. Genomewide association analysis identified peaks on equine A) chromosomeWalking Manhattan (ECA) Horse. plot 19 of and for the ECA11 Tennessee gait permutated type –log variation in B) The 1.4-Mb region surrounding the candidate markers on ECA11 spans (C) several candidate genes in (D) 2 blocks of linkage disequilibrium (LD). E) The 1.3-Mb region surrounding the candidate markers on ECA19 markers candidate surrounding the 1.3-Mb region The (LD). E) disequilibrium genes in (D) 2 blocks of linkage candidate spans (C) several on ECA11 markers candidate surrounding the 1.4-Mb region The B) spans (F) several candidate genes in (G) a block large of LD. Single nucleotide polymorphisms outlined with black boxes in the Haploview (version 4.2; Barrett et al., 2005) panels (D and G) are the significant haplotype http://genome.ucsc.edu (accessed December 31, 2015; Kent et al., 2002). analysis blocks. Gene track images from the University of California Santa Cruz genome browser, 1382 Staiger et al. Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021

Figure 3. Tennessee Walking Horse (TWH) population structure and linkage disequilibrium (LD) analysis. A) Genomewide association multidimensional scaling analysis dimensions 1 versus 3. Red dots represent lateral-only gaited horses and blue dots represent multigaited horses. B) Average genomewide LD in 4 different breeds, with 2 subpopulations from 1 breed. The TWH LD is calculated from genotypes generated on the Illumina SNP70 beadchip (Illumina Inc., San Diego, CA); all other breeds are genotyped on the Illumina SNP50 beadchip (Illumina Inc.).

23.5%) and 8% (range 0.30 to 23.4%), respectively, at a low frequency in this breed. Nine of the horses and were not different between the 2 groups (Students categorized by owner observations may have been t test, P = 0.92). The 2 individuals without pedigrees misclassified as able to trot due to the inexperience of possessed low genomic estimates of inbreeding ( pi some owners; we sampled horses from a broad range values of 0.002744 and 0.003671) compared with the of owner experience level. sample mean (0.037314; range 0.792 to 0). Variation in horse age (1 to 34 yr) and, there- Ten unrelated individuals from this study as well as fore, training and experience level also introduces an additional 10 unrelated individuals from the Egyptian error in our phenotyping scheme. Older horses are Arabian, Thoroughbred, Arabian (non-Egyptian), and more likely to have more training and are often better American Saddlebred breeds were used to calculate conditioned to perform one end of the gait spectrum. average genomewide LD (Fig. 3b). Linkage disequilib- Horses trained to stay in one gait type over another rium in the TWH is shorter than that of the Egyptian are likely to have adapted muscle conformations and Arabian and Thoroughbred but is most similar to the muscle memory for that specific gait, because skeletal Arabian (non-Egyptian) and the American Saddlebred. muscle properties are dynamic and reflect the history of use (Snow and Valberg, 1994; Rivero, 2007; DISCUSSION Riley and Van Dyke, 2012) and differences in muscle conformation and mass can influence how the horses We identified several genomic regions that are sug- perform their gaits (Ziegler, 2005). Horses with ex- gestively associated with gait type in the TWH and tensive training are typically used for showing, which, warrant further investigation in larger populations. in turn, means they are also more likely to wear as- Additional research in a larger sample size, using a sociated equipment, such as heavier shoes, to enhance denser SNP array and using a more quantitative means the lateral end of the gait spectrum and discourage of measuring the phenotype, would improve the power the use of the trot by increasing the swing phase of of association and likely confirm our identified regions. the limb (Clayton, 2004). Shoeing with weights and The intermediate gaits of the TWH are a continu- wedges (as observed in light shod, plantation, and per- ous spectrum that blend together and fluctuate based formance show horses) also influences tendon tension on speed and terrain as well as the influence and skill by reducing the tension, prolonging break-over of the level of the rider. In our study, we attempted to pheno- stride, and changing the interaction between tendons type based on owner-reported performance or a short and muscles (Riemersma et al., 1996; Biewener, 1998; video observation of the horse moving in the interme- Clayton, 2004; Lawson et al., 2007), thereby chang- diate gaits and then grouped horses based on presence/ ing the mechanics of the gait. Older horses with lon- absence of diagonal pair footfall landings. All horses ger training histories could have greater muscle mass within the TWH study were able to use lateral pair and, therefore, be more conditioned to perform one footfall landings, indicating that alleles conferring a gait type over another. However, it is important to lingering ability to use a diagonal footfall pattern are note that age did not significantly differ between the Genetics of gait in Tennessee Walking horses 1383

2 gait types in our study (Supplemental Fig. S3; see pacity for motor neuron signaling, longer axon lengths, the online version of the article at http://journalofani- or different neuronal growth patterns. malscience.org). Improving phenotype ascertainment Our ECA19 candidate region encompasses 2 dif- could overcome difficulties encountered in this study. ferent but functionally interesting genes. The FBXO40 The use of accelerometers has proven useful in gait gene is a member of the F-box protein family that are analysis (Witte et al., 2004; Pfau et al., 2005, 2006; characterized as components of SCF (Skp1-Cullin1-F- Robilliard et al., 2007) and in distinguishing lateral box protein) E3 ubiquitin ligase complexes, in which from diagonal 4-beat gaits (Östlund, 2011). The ac- they bind substrates for ubiquitin-mediated proteolysis celerometers can be used to determine stride kine- (Deshaies, 1999; Cardozo and Pagano, 2004; Lin and matics, speed, and, potentially, head motion (head Diehl, 2004; Ye et al., 2007). The gene is expressed in motion is associated with the running walk and fox skeletal muscle and functions as a regulator involved trot). Equipment consistently measuring the speed of in postnatal myogenesis (Ye et al., 2007) and regulat- the gait would allow for a more uniform comparison ing IGF-1 (Shi et al., 2011). Muscles initiate and con- Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 between horses moving at the same speed. In addition, trol movement through innervation by motor neurons, the maximum speed before a horse breaks gait into the which originate in the spinal cord but receive projec- canter and kinematics of the canter could be measured tions from neurons originating from the brainstem and tested for association to the DMRT3 mutation. and midbrain (Yuste et al., 2005). Differences in the Due to the continuous nature of the various types muscles of gaited horses could be altering the signaling of intermediate gaits, interaction of multiple genes as pathways of the neurons involved in the CPG pathways. well as multiple alleles that are pleiotropic are likely The ARGFX (arginine-fifty homeobox) gene is a genetic mechanisms for this trait. For this reason, we member of the homeobox genes that encode DNA- applied several different analysis methods to our data binding proteins, many of which are thought to be in- set. First, we applied an allelic association to check if volved in early embryonic development. Interestingly, there was perhaps a single gene with a major effect on an open reading frame for ARGFX has been identified the phenotype. When the analysis failed to identify any only in the human genome but not in mice, horses, markers surpassing a Bonferroni correction, we next cows, or guinea pigs and cannot be distinguished as ei- applied a basic adaptive permutation test. One marker ther a true functional gene or a nonfunctional pseudo- surpassed our significance threshold of 1 million per- gene (Li and Holland, 2010). The gene likely originat- mutations, but the P-values were similar to the allelic ed by from a gene duplication of OTX1, OTX2, or CRX association. Genomic inflation and MDS analysis did during early mammalian evolution (Li and Holland, not detect any population substructure between cases 2010). OTX1 is a member of the homeodomain-con- and controls, yet any undetected confounding popula- taining transcription factors and, in the mouse, is re- tion structure that may still exist in the data set is not quired for proper brain and sensory organ development preserved following the shuffling of phenotypes across (Booth and Holland, 2007). OTX2 is also a member genotypes during permutation (Che et al., 2014). To in- of the homeodomain-containing transcription factors crease our statistical power, we next applied haplotype and plays a role in brain, craniofacial, and sensory or- analysis to the regions surpassing 30,000 permutations. gan development (Booth and Holland, 2007). CRX is a Two regions from the permutation were confirmed by photoreceptor-specific transcription factor that plays a this method. Several of the genes we identified were role in the differentiation of photoreceptor cells (Booth related to immunology, whereas there are a dispropor- and Holland, 2007; Li and Holland, 2010). tionally large number of immune-related genes; there The SMTNL2 gene falls within our ECA11 candi- is the potential that we could have either inadvertently date region and has been identified as a new substrate run an association on disease status or we have identi- for mitogen-activated protein kinases (MAPK), which fied pleiotropic genes that play important roles in gait execute several critical cell processes, including cell type and complex disease susceptibility. division, differentiation, and stress response (Gordon Despite these challenges, this study highlights et al., 2013). SMTNL2 is expressed in several mamma- some suggestive candidate regions containing genes lian tissues but shows the highest expression in skel- that may prove to play a role in modulation of this trait. etal muscle and appears to be involved in myogenic Several of these regions include genes involved in de- differentiation (Gordon et al., 2013). The rigidity and velopmental process, responses to stimuli, and biolog- flexibility of the horse’s body frame can influence ical regulation (Supplemental Fig. S4; see the online which gait is performed (Ziegler, 2005); a genetic dif- version of the article at http://journalofanimalscience. ference in muscle composition can predispose a horse org). It may prove that gaited horses have greater ca- toward a body frame that is easier to maintain and, therefore, a preferred gait type. 1384 Staiger et al. Tennessee Walking Horse Population Structure Despite their polymorphic gait phenotype, the TWH is nearly fixed for the DMRT3 stop mutation Linkage disequilibrium length was shortest in the previously reported by Andersson et al. (2012). In the Saddlebred, TWH, and Arabian, with r2 values drop- DMRT3 mouse knockout, longer strides were observed ping below 0.5 within the first 10 to 20 kb. Linkage as speed increased, but the mouse was unable to gallop disequilibrium was the longest in the Thoroughbred (Andersson et al., 2012). Anecdotally, this is a com- and Egyptian Arabians, which did not drop below 0.5 mon feature observed in the TWH. Although speed r2 and reflects the breeds’ high inbreeding, low diver- was not measured in our gait analysis, the TWH can sity, and closure of the studbooks to outside genetic travel 16 to 32 km/h at an intermediate gait (Tennessee influence for several hundred years. The Saddlebred Walking Horse Breeders’ and Exhibitors’ Association, breed registry was closed in 1917 and contributed to 2011) rather than using the canter, which has a typi- the TWH breed registry, which itself was closed in cal speed of 16 to 27 km/h (Clayton, 1993, 1994). 1947. The longer extent of LD in the TWH compared Harness racing horses reach speeds around 40 to 48 with the Saddlebred and Arabian may reflect a higher Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 km/h (Barrey, 2013) at both the trot and pace, and yet level of inbreeding or a smaller founding population these horses were also observed to have the DMRT3 in the TWH than the latter 2 breeds. Higher individual mutation at high frequency (Andersson et al., 2012). inbreeding levels are observed in the TWH compared Even though previous studies have shown an associa- with the Arabian and Saddlebred (F mean = 0.148, F tion between the mutation and the ability to perform mean = 0.060, and F mean = 0.103, respectively) by alternate gaits, the frequency of DMRT3 mutation was McCue et al. (2012). This can be explained by fre- detected in a wide distribution of breeds, gaited and quent use of line breeding (Bourdon, 2000), which has nongaited (Promerová et al., 2014), indicating DMRT3 occurred in the TWH since its foundation (Womack, is more likely associated with speed and transitions 1994). Indeed, every horse registered today with the from one gait pattern to the next. We hypothesize that Tennessee Walking Horse Breeders’ and Exhibitors’ the DMRT3 mutation is not permissive for “gaitedness” Association can trace back to 1 stallion born in 1940, but plays an inhibitory role in switching from interme- Midnight Sun. The individuals in this study with the diate gaits into the canter. Because many of the gaited highest inbreeding values traced their ancestry back to breeds are able to perform their alternate 4-beat gaits Midnight Sun multiple times in the pedigree. at higher speeds equivalent to a canter, it is likely that Typically, the modern TWH is classified into 3 in selecting for lateral gaits, there was simultaneous major stallion lines descending from Midnight Sun. selection against use of the canter. Additionally, the Two of the lines are grandsons from the same son of presence of this mutation in other, nongaited breeds Midnight Sun, Pride of Midnight, who was bred to indicates there must be another region controlling the daughters of his half-brothers (Spirit of Midnight and ability to perform a 4-beat gait. Sun’s Delight D). The third is a great-grandson who Indeed, selection sweeps across other breeds have is blended with another popular sire during Midnight identified additional loci beyond ECA23 (Petersen et al., Sun’s lifespan, Merry Go Boy. These lines are likely 2013). The TWH was included in Petersen’s work and what have been detected in the comparison of the di- suggested selection signals on ECA8, ECA16, ECA19, mensions 1 and 3 in the MDS plots, based on the ex- ECA23, and ECA27 (Petersen et al., 2013). Our anal- amination of the pedigree tree for the study horses. ysis also included markers from ECA8 and ECA19, lending additional support to our identified candidate Conclusions regions. However, it is important to note that the previous study used a different approach. In the Petersen Due to the complex nature of locomotion, with study, only 17 horses were used in a comparison of the its continuous phenotype and multiple environmen- TWH breed with 32 other breeds. Therefore, given that tal factors, genetic mapping of gait genes is a diffi- the breed is nearly fixed for the ECA23 stop mutation cult task. At a rudimentary level of phenotyping, we marker, it is not surprising that the locus appeared in the demonstrate with the current genotyping technology selection sweep. Additionally, all 17 TWH used were that there are detectable haplotype differences within assumed to have the same gait phenotype, although the the TWH breed. Future work will pursue genotyping a gait of these animals was not observed and verified. Our larger set of horses and/or markers to improve detec- study was a within-breed comparison and documented tion of multiple regions with small effects contribut- all phenotypes by video recording. ing to this gait phenotype. The identified markers can be applied for prediction of gait type, aiding breeders in their selection and breeding plans and improving the marketability of their horses. These tests could Genetics of gait in Tennessee Walking horses 1385

Coonan, J. R., U. Greferath, J. Messenger, L. Hartley, M. Murphy, also decrease the amount of time and money spent A. W. Boyd, M. Dottori, M. P. Galea, and P. F. Bartlett. 2001. on training; training schemes can be personalized to Development and reorganization of corticospinal projec- each individual horse’s genetic potential and ineffec- tions in EphA4 deficient mice. J. Comp. Neurol. 436:248–262. tive techniques can be avoided. Animal models have doi:10.1002/cne.1064 long played an important role in human disease re- Crone, S. A., K. A. Quinlan, L. Zagoraiou, S. Droho, C. E. Restrepo, L. Lundfald, T. Endo, J. Setlak, T. M. Jessell, O. Kiehn, and K. search due to similarities in basic biology, physiology, Sharma. 2008. Genetic ablation of V2a ipsilateral interneurons and experimental convenience. By understanding the disrupts left-right locomotor coordination in mammalian spinal different gait types in the horse, new insights can be cord. Neuron 60:70–83. doi:10.1016/j.neuron.2008.08.009 provided on motor coordination, as no other mamma- Deshaies, R. J. 1999. SCF and Cullin/Ring H2-based ubiquitin li- lian species is known to discretely segregate for lateral gases. Annu. Rev. Cell Dev. Biol. 15:435–467. doi:10.1146/annurev.cellbio.15.1.435 stride cadence and footfall pattern. Fletcher, J. L. 1946. A study of the first fifty years of Tennessee Walking Horse breeding. J. Hered. 37:369–373. Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 Literature Cited Garbe, J. R., and Y. Da. 2008. Pedigraph: A software tool for the graphing and analysis of large complex pedigree. User manual. Albertsdóttir, E., S. Eriksson, A. Näsholm, E. Strandberg, and Ver. 2.4. Dep. Anim. Sci., Univ. Minnesota, St. Paul. MN. T. Árnason. 2008. Genetic correlations between competition Gordon, E. A., T. C. Whisenant, M. Zeller, R. M. Kaake, W. M. traits and traits scored at breeding field-tests in Icelandic horses. Gordon, P. Krotee, V. Patel, L. Huang, P. Baldi, and L. Bardwell. Livest. Sci. 114:181–187. doi:10.1016/j.livsci.2007.04.022 2013. Combining docking site and phosphosite predictions Andersson, L. S., M. Larhammar, F. Memic, H. Wootz, D. to find new substrates: Identification of smoothelin-like-2 Schwochow, C. J. Rubin, K. Patra, T. Arnason, L. Wellbring, G. (SMTNL2) as a c-Jun N-terminal kinase (JNK) substrate. Cell. Hjalm, F. Imsland, J. L. Petersen, M. E. McCue, J. R. Mickelson, Signal. 25:2518–2529. doi:10.1016/j.cellsig.2013.08.004 G. Cothran, N. Ahituv, L. Roepstorff, S. Mikko, A. Vallstedt, G. Grillner, S., and P. Zangger. 1979. On the central generation of Lindgren, L. Andersson, and K. Kullander. 2012. Mutations in locomotion in the low spinal cat. Exp. Brain Res. 34:241–261. DMRT3 affect locomotion in horses and spinal circuit function doi:10.1007/BF00235671 in mice. Nature 488:642–646. doi:10.1038/nature11399 Harris, S. E. 1993. Horse gaits, balance, and movement. Howell Barrett, J. C., B. Fry, J. Maller, and M. J. Daly. 2005. Haploview: Book House, New York, NY. Analysis and visualization of LD and haplotype maps. Hinckley, C. A., R. Hartley, L. Wu, A. Todd, and L. Ziskind- Bioinformatics 21:263–265. doi:10.1093/bioinformatics/bth457 Conhaim. 2005. Locomotor-like rhythms in a genetically dis- Barrey, E. 2013. Gaits and interlimb coordination. In: W. Back and tinct cluster of interneurons in the mammalian spinal cord. J. H. M. Clayton, editors, Gaits and interlimb coordination. 2nd ed. Neurophysiol. 93:1439–1449. doi:10.1152/jn.00647.2004 Elsevier, New York, NY. p. 85–97. Kent, W. J., C. W. Sugnet, T. S. Furey, K. M. Roskin, T. H. Pringle, Biewener, A. A. 1998. Muscle-tendon stresses and elastic en- A. M. Zahler, and D. Haussler. 2002. The human genome brows- ergy storage during locomotion in the horse. Comp. Biochem. er at UCSC. Genome Res. 12:996-1006. Physiol. B Biochem. Mol. Biol. 120:73–87. doi:10.1016/S0305- Kiehn, O. 2006. Locomotor circuits in the mammalian spinal cord. 0491(98)00024-8 Annu. Rev. Neurosci. 29:279–306. doi:10.1146/annurev.neu- Booth, H. A. F., and P. W. H. Holland. 2007. Annotation, no- ro.29.051605.112910 menclature and evolution of four novel homeobox genes ex- Kullander, K., S. D. Croll, M. Zimmer, L. Pan, J. McClain, V. Hughes, pressed in the human germ line. Gene 387:7–14. doi:10.1016/j. S. Zabski, T. M. DeChiara, R. Klein, G. D. Yancopoulos, and N. gene.2006.07.034 W. Gale. 2001. Ephrin-B3 is the midline barrier that prevents Bourdon, R. M. 2000. Mating strategies based on pedigree relation- corticospinal tract axons from recrossing, allowing for unilateral ship: Inbreeding and outbreeding. 2nd ed. Prentice Hall, Upper motor control. Genes Dev. 15:877–888. doi:10.1101/gad.868901 Saddle River, NJ. p. 333–370. Kuo, A. D. 2002. The relative roles of feedforward and feedback Cardozo, T., and M. Pagano. 2004. The SCF ubiquitin ligase: in the control of rhythmic movements. Mot. Contr. 6:129–145. Insights into a molecular machine. Nat. Rev. Mol. Cell Biol. Lawson, S. E. M., H. Chateau, P. Pourcelot, J. Denoix, and N. 5:739–751. doi:10.1038/nrm1471 Crevier-Denoix. 2007. Effect of toe and heel elevation on calcu- Che, R., J. R. Jack, A. A. Motsinger-Reif, and C. C. Brown. 2014. An lated tendon strains in the horse and the influence of the proximal adaptive permutation approach for genome-wide association study: interphalangeal joint. J. Anat. 210:583–591. doi:10.1111/j.1469- Evaluation and recommendations for use. BioData Min. 7:9. 7580.2007.00714.x Clayton, H. M. 1993. The extended canter: A comparison of some Li, G., and P. W. H. Holland. 2010. The origin and evolution of kinematic variables in horses trained for dressage and for racing. ARGFX homeobox loci in mammalian radiation. BMC Evol. Acta Anat. (Basel) 146:183–187. doi:10.1159/000147443 Biol. 10:182. doi:10.1186/1471-2148-10-182 Clayton, H. M. 1994. Comparison of the collected, working, me- Lin, D. I., and J. A. Diehl. 2004. Mechanism of cell-cycle con- dium and extended canters. Equine Vet. J. Suppl. 17:16–19. trol: Ligating the ligase. Trends Biochem. Sci. 29:453–455. Clayton, H. M. 2004. The dynamic horse: A biomechanical guide doi:10.1016/j.tibs.2004.07.006 to equine movement and performance. 1st ed. Sport Horse Publications, Mason, MI. Cook, D., P. C. Gallagher, and E. Bailey. 2010. Genetics of sway- back in American Saddlebred horses. Anim. Genet. 41:64–71. doi:10.1111/j.1365-2052.2010.02108.x 1386 Staiger et al.

McCue, M. E., D. L. Bannasch, J. L. Petersen, J. Gurr, E. Bailey, Riemersma, D. J., A. J. van der Bogert, M. O. Jansen, and H. C. M. M. Binns, O. Distl, G. Guerin, T. Hasegawa, E. W. Hill, T. Schamhardt. 1996. Influence of shoeing on ground reaction Leeb, G. Lindgren, M. C. Penedo, K. H. Roed, O. A. Ryder, J. E. forces and tendon strain in the forelimbs of ponies. Equine Vet. J. Swinburne, T. Tozaki, S. J. Valberg, M. Vaudin, K. Lindblad-Toh, 28:126–132. doi:10.1111/j.2042-3306.1996.tb01604.x C. M. Wade, and J. R. Mickelson. 2012. A high density SNP ar- Riley, D. A., and J. M. Van Dyke. 2012. The effects of active and ray for the domestic horse and extant Perissodactyla: Utility for passive stretching on muscle length. Phys. Med. Rehabil. Clin. association mapping, genetic diversity, and phylogeny studies. N. Am. 23:51–57. doi:10.1016/j.pmr.2011.11.006 PLoS Genet. 8:e1002451. doi:10.1371/journal.pgen.1002451 Rivero, J.-L. L. 2007. A scientific background for skeletal muscle Östlund, V. 2011. Limb phasing Icelandic horses. Swedish conditioning in equine practice. J. Vet. Med., A 54:321–332. University of Agricultural Sciences, Uppsala, Sweden. Master’s doi:10.1111/j.1439-0442.2007.00947.x thesis. Swedish University of Agricultural Sciences, Uppsala, Robilliard, J. J., T. Pfau, and A. M. Wilso. 2007. Gait characteri- Sweden. sation and classification in horses. J. Exp. Biol. 210:187–197. Patterson, L., E. A. Staiger, and S. A. Brooks. 2015. DMRT3 is asso- doi:10.1242/jeb.02611 ciated with gait type in Mangalarga Marchador horses, but does Shi, J., L. Luo, J. Eash, C. Ibebunjo, and D. J. Glass. 2011. The SCF- not control gait ability. Anim. Genet. 46:213–215. doi:10.1111/ Fbxo40 complex induces IRS1 ubiquitination in skeletal muscle, Downloaded from https://academic.oup.com/jas/article/94/4/1377/4703958 by guest on 29 September 2021 age.12273 limiting IGF1 signaling. Dev. Cell 21:835–847. doi:10.1016/j. Petersen, J. L., J. R. Mickelson, A. K. Rendahl, S. J. Valberg, L. S. devcel.2011.09.011 Andersson, J. Axelsson, E. Bailey, D. L. Bannasch, M. M. Binns, Snow, D. H., and S. J. Valberg. 1994. Muscle anatomy, physiology A. S. Borges, P. Brama, A. da Camara Machado, S. Capomaccio, and adaptations to exercise and training. In: D. R. Hodgson and K. Cappelli, E. G. Cothran, O. Distl, L. Y. Fox-Clipsham, K. R. J. Rose, editors, The Athletic Horse: Principles and Practice of T. Graves, G. Guerin, B. Haase, T. Hasegawa, K. Hemmann, Equine Sports Medicine. WB Saunders Company, Philadelphia, E. W. H. Ill, T. Leeb, G. Lindgren, H. Lohi, M. S. Lopes, B. A. PA. p. 145–179. McGivney, S. Mikko, N. Orr, M. C. Penedo, R. J. Piercy, M. Tennessee Walking Horse Breeders’ and Exhibitors’ Association. Raekallio, S. Rieder, K. H. Roed, J. Swinburne, T. Tozaki, M. 2011. The Tennessee Walking Horse breed: Gaits. http://www. Vaudin, C. M. Wade, and M. E. McCue. 2013. Genome-wide twhbea.com/breed/gait.php. Accessed November 13, 2014. analysis reveals selection for important traits in domestic Wilson, J. M., R. Hartley, D. J. Maxwell, A. J. Todd, I. Lieberam, J. horse breeds. PLoS Genet. 9:e1003211. doi:10.1371/journal. A. Kaltschmidt, Y. Yoshida, T. M. Jessell, and R. M. Brownstone. pgen.1003211 2005. Conditional rhythmicity of ventral spinal interneurons de- Pfau, T., T. H. Witte, and A. M. Wilson. 2005. A method for deriving fined by expression of the Hb9 homeodomain protein. J. Neurosci. displacement data during cyclical movement using an inertial 25:5710–5719. doi:10.1523/JNEUROSCI.0274-05.2005 sensor. J. Exp. Biol. 208:2503–2514. doi:10.1242/jeb.01658 Witte, T. H., K. Knill, and A. M. Wilson. 2004. Determination of Pfau, T., T. H. Witte, and A. M. Wilson. 2006. Centre of mass move- peak vertical ground reaction force from duty factor in the horse ment and mechanical energy fluctuation during gallop locomo- (Equus caballus). J. Exp. Biol. 207:3639–3648. doi:10.1242/ tion in the Thoroughbred racehorse. J. Exp. Biol. 209:3742– jeb.01182 3757. doi:10.1242/jeb.02439 Womack, B. 1994. The echo of hoofbeats: A history of the Tennessee Promerová, M., L. S. Andersson, R. Juras, M. C. T. Penedo, M. Walking Horse. 3rd ed. Dabora, Inc., Shelbyville, TN. Reissmann, T. Tozaki, R. Bellone, S. Dunner, P. Horín, F. Imsland, Ye, J., Y. Zhang, J. Xu, Q. Zhang, and D. Zhu. 2007. FBXO40, a P. Imsland, S. Mikko, D. Modrý, K. H. Roed, D. Schwochow, J. gene encoding a novel muscle-specific F-box protein, is upregu- L. Vega-Pla, H. Mehrabani-Yeganeh, N. Yousefi-Mashouf, E. G. lated in denervation-related muscle atrophy. Gene 404:53–60. Cothran, G. Lindgren, and L. Andersson. 2014. Worldwide fre- doi:10.1016/j.gene.2007.08.020 quency distribution of the “Gait keeper” mutation in the DMRT3 Yuste, R., J. N. MacLean, J. Smith, and A. Lansner. 2005. Opinion: gene. Anim. Genet. 45:274–282. doi:10.1111/age.12120 The cortex as a central pattern generator. Nat. Rev. Neurosci. Purcell, S., B. Neale, K. Todd-Brown, L. Thomas, M. A. Ferreira, 6:477–483. doi:10.1038/nrn1686 D. Bender, J. Maller, P. Sklar, P. I. de Bakker, M. J. Daly, and P. Ziegler, L. 2005. Easy-gaited horses. 1st ed. Storey Publishing, C. Sham. 2007. PLINK: A tool set for whole-genome associa- North Adams, MA. tion and population-based linkage analyses. Am. J. Hum. Genet. 81:559–575. doi:10.1086/519795